Recent research in active learning, and more precisely in uncertainty sampling, has focused on the decomposition of model uncertainty into reducible and irreducible uncertainties. In this paper, we propose to simplify the computational phase and remove the dependence on observations, but more importantly to take into account the uncertainty already present in the labels, i.e. the uncertainty of the oracles. The proposed strategy, sampling by Klir uncertainty, addresses the exploration-exploitation dilemma using the theory of belief functions.